What Is a Certificate of Analysis? Specification vs Result

A certificate of analysis is a document reporting what a laboratory found when it tested a specific production lot of a specific material.

That definition is doing a lot of work, and most of what circulates in this category under the name does not meet it. The difference is one column, and it is the most useful thing anyone can learn about supplement paperwork.

What is a certificate of analysis?

A report of test results on a named lot of material. A complete one names the material, the lot, the date, each test performed, the method used, the specification each test was measured against, and the result the test returned. It is issued by whoever ran or commissioned the testing, and it is signed.

Read that list again and count how many of those items appear on a document a supplement brand has sent you. The gap between the definition and the shelf is where this entire subject lives.

What is the difference between a specification and a result?

A specification is what a formula is supposed to be. A result is what a test found. A specification sheet lists limits — lead below three parts per million, identity above 95% — and stops. A certificate of analysis lists those same limits and then prints, beside each one, the number that came back.

Both documents are useful and only one of them is a measurement. A specification-only sheet describes an intention. A result describes an event. The category calls them both certificates, which is how a document that measured nothing ends up being cited as proof.

How do you tell a limit from a result?

A result is a number an instrument returned. A limit is a number somebody chose in advance. They sit in adjacent columns, they are set in the same typeface, and only one of them is evidence. On a heavy-metal page three of the numbers are never results: the limit of quantification, the limit of detection and the acceptance limit — the smallest quantity the method can report, the smallest it can detect at all, and the ceiling the rule allows. None of the three has anything to do with the sample.

Night Work’s certificate is the clearest illustration of this we can offer, because it is our own and because its heavy-metal page carries all four kinds of number side by side. It is set out below under the document’s own column headings.

Every figure in that table is real and only four of them are measurements of this powder. Beside lead, 0.044 is what the instrument returned; 0.010 is the smallest quantity the method could have reported, 0.005 the smallest it could have detected, and 5 the daily ceiling the rule sets. Cadmium and mercury read <LOQ, below 0.010 ppm — a floor the method stops at, not a quantity. And the one column that carries a finding is in parts per million while the Limit column is in micrograms per day, with no conversion printed between them, so the Pass is the laboratory’s verdict and no margin is ours to compute. The four readings are the finding — three of this document’s four axes carry one and identity does not, which is why Night Work publishes at a Lot Mark of three of four.

What does “Serving Size: N/A” on a certificate tell you?

That no serving was tested, because the sample never had one. A filled tub or a finished capsule has a serving size; a sack of raw material does not. When that field reads N/A the document is describing an ingredient before it became a product, and every figure on it is a figure about the ingredient rather than about anything a person could buy. It is the fastest question on the page to answer and it changes the meaning of everything printed below it.

It appears on one of the two certificates TARE holds from an outside laboratory. Base Work’s names its sample “12335 - Creatine Monohydrate”, records the sample type as Powder and the serving size as N/A, and photographs a sealed poly bag marked Accobio USA Inc., lot CM20251225A. Night Work’s names its sample “FG - 13072 - Magnesium Glycinate”, records the sample type as Powder and prints a serving size of 926 mg — the laboratory’s own serving, which the document never relates to a capsule, so every figure on it is still a figure about a sack of powder — and photographs a sealed poly bag labelled Magnesium Glycinate carrying the sublot line MGL260212-01-0001, 505 kilograms, and the word Lab. written on it by hand, which the document does not explain.

Who is the customer named on a certificate, and why is it not the brand?

The customer of record is whoever commissioned the work and paid for it, and it is printed on the face of the document. It is not necessarily the brand handing you the certificate. On both of the documents TARE holds from an outside laboratory the customer is Vox Nutrition, the ingredient supplier — not TARE, and not the company that manufactures for TARE. A brand quoting a certificate it did not commission is quoting somebody else’s paperwork, which is not the same thing as dishonest and is not the same thing as its own.

On Base Work’s document the customer is printed as Vox Nutrition, 8224 Industry Circle, West Jordan UT 84088. On Night Work’s the header field prints the name alone; the footer reads “Ethos Analytics Tested for Vox Nutrition:” above the supplier’s own logo, and the line beneath it marked “Vox Nutrition Quality Sign Off:” is blank. Both are signed by Noel Samsum, Laboratory Director — 18 February 2026 for the creatine, 24 February 2026 for the magnesium.

Both documents also close off the use a brand would most like to make of them, in the laboratory’s own footer: “The presented report is only applicable to the sample specified above and may not be applied to any similar or identical products.” Which leaves the sentence that costs us something to print: two of the three certificates on this site name raw-material lots bought for another company, and only one of the three names the lot inside a TARE tub.

What is an identity test, and why does it matter more than potency?

An identity test asks whether the material is the substance the label names. A potency test asks how much of it is there. Identity has to come first, because a potency figure inherits whatever identity established — measure the wrong compound accurately and the result is an accurate number about the wrong thing. Identity is usually an infrared scan or a chromatographic fingerprint read against a reference, it is the cheapest of the four axes to run, and it is the one most often left off.

On Night Work’s certificate identity was NOT RUN. There is no identity section on the document at all, which means nothing on it separates magnesium glycinate from magnesium oxide, magnesium citrate or a blend of several; its magnesium assay counts the element by ICP-MS without saying which salt carried it. On Under Load’s certificate identity was NOT RUN as well: the ingredient line reads “collagen peptides”, which is a statement written on a form rather than a test performed on a sample. Base Work’s is the only one of the three carrying a figure that speaks to identity, and it does so by assay rather than by an identity method — creatine monohydrate at 99.87% against a specification of more than 99.5%, by HPLC. TARE counts that row as identity as well as potency and prints the objection to counting it twice on the certificate reference.

What is heavy-metal testing, and what is Proposition 65?

Heavy-metal testing measures how much lead, arsenic, cadmium and mercury a material carries, usually by ICP-MS. Proposition 65 is a Californian law setting a daily exposure figure above which a product sold in that state has to carry a warning. It is a labelling threshold rather than a purity grade, so a figure under it says a warning is not required and says nothing else. It also sits on a different axis from the result beside it: the result is per serving of the material tested, and the Proposition 65 figure is per day.

Under Load’s certificate reports lead at 0.07 mcg against a 0.5 mcg Proposition 65 limit, with mercury, arsenic and cadmium below the detectable limit, by ICP-MS on lot 121125450 — a 350 g lot of the collagen a TARE tub is filled from. Base Work’s reports ICP-MS on a numeric USP <233> page, every line a pass, on the ingredient supplier’s poly bag. Night Work’s reports ICP-MS under USP <233> on a raw-material lot for the same supplier: arsenic 0.034 ppm, lead 0.044 ppm, cadmium and mercury below the limit of quantification, every row a pass, as set out above.

What does a microbiology pass actually cover?

Counts and presence tests, and nothing beyond them. A microbiology page counts aerobic organisms, coliforms, yeast and mould per gram against a ceiling, and tests whether E. coli, salmonella and staphylococcus aureus can be found in a ten-gram sample. A pass means those organisms were under their ceilings or absent on the day the sample was tested. It says nothing about what the material is, nothing about how much active it carries, and nothing about metals.

The same Night Work document that measured its metals and its magnesium measured all seven organisms, which makes it the plainest illustration available to us: two pages on which every organism and every metal passes, and not one line on either that says what the powder is.

Run on 23 February 2026 by Petrifilm and qPCR, every row a pass. Two readings are worth taking from it. A result of under 20 colony-forming units per gram is not a count of zero — it is the point below which the method stops resolving, which is the same lesson as the limit of quantification one table up. And the honest description of the whole document, in the only words we are willing to use for it: the magnesium glycinate raw material was tested by Ethos Analytics, for Vox Nutrition, on lot MGL260212, completed 24 February 2026, for magnesium, heavy metals and microbiology: 102.54 mg of magnesium per 926 mg laboratory serving against a 91.6 mg specification, heavy metals pass, microbiology pass. That certificate carries no identity test, its serving is the laboratory’s and not the label’s three capsules, and it does not test a TARE container.

What should be checked on a certificate?

Eight things, in this order. Whether there is a result column or only a limits column. Whether a lot number is printed. Whether the sample was a finished product or a raw material. Whether the serving-size field carries a serving. Who the customer of record is. Whether a method is named beside each test. Whether it carries a date and a signature. And whether it says what it did not test.

1. Is there a result column, or only a limits column? This decides everything else.
2. Is a lot number printed? A document with no lot is not about any particular material.
3. Was the sample a finished product or a raw material? Both produce real numbers. Only one of them was ever a product.
4. Does the serving-size field carry a serving? N/A answers the question above it, and answers it against the brand.
5. Who is the customer of record? A certificate is addressed to whoever commissioned it. If that is not the brand selling to you, the brand did not commission it.
6. Is a method named beside each test? A percentage without a method is not comparable to another percentage.
7. Is there a date and a signature? A test date, a document date, and a named person standing behind both.
8. Does it say what it did not test? The strongest documents do. Almost none do.

Two companion lists exist and neither is this one. Four questions settle which of the three jobs the word “tested” is doing on a given badge, and six are put to TARE’s own three documents with the answers printed beside them.

What do TARE's own certificates report?

All three report measured results — a number printed beside the specification it was measured against. Night Work's does so in the laboratory's own serving, on a raw-material lot, with no identity test. That is a change and it is worth saying plainly: until August 2026 we published four certificates and two of them reported no test result of any kind — and both of those products are now retired. Base Work, lot CM20251225A, reports creatine monohydrate at 99.87% against a greater-than-99.5% specification, by HPLC. Under Load, lot 121125450, reports protein at 9.8 g against a 9 g specification, and lead at 0.07 mcg against a 0.5 mcg Proposition 65 limit — but those two figures are not on the same axis. The result is per 10 g serving; the Proposition 65 figure is a per-day limit. At the 15 g protocol that is 0.105 mcg against 0.5 mcg, 4.8 times under, and that is the margin we publish. Night Work’s document, lot MGL260212, reports magnesium at 102.54 mg against a 91.6 mg specification by ICP-MS — in the laboratory’s 926 mg serving rather than in three capsules — with a heavy-metal page and a microbiological page beside it, and no identity test. The figure is the powder’s, not the capsule’s, which is exactly what a Lot Mark of three of four with identity not run records.

One of those is the strongest thing this company owns and it is worth spelling out. The collagen protein figure is larger than the specification, not equal to it — 9.8 g found against 9 g promised. Run it forward to the 15 g protocol dose and the specification is 13.5 g while the laboratory’s 9.8 g scales to 14.7 g — arithmetic, not a second test. We claim less than the document found, and we publish both numbers. That is the only direction that arithmetic is allowed to run here.

What replaces the old two-weak-documents caveat is a different split, and it is not a softer one. All three carry a measured result; they were not all run by the same kind of laboratory, and only one names a TARE lot, which is the subject of the next section. Every row this site argues from is transcribed on Testing & Certification.

Where our own documents stop

A laboratory is named on all three, and they are not the same kind of laboratory. Base Work’s creatine and Night Work’s magnesium were run by an outside laboratory in Phoenix, Arizona that had nothing to do with making either product. Under Load’s collagen was run by the company that manufactures it. Both are real documents with real numbers in them, and the second is a weaker check than the first — a maker grading its own work — which the reader should hear from us rather than work out later. It is our hero product and it is the one we would tell you to buy last on documents alone. Which laboratory ran which is on Testing & Certification, at the top of each panel.

Every certificate names the lot it was run on, and it is not the reader’s lot. That distinction is not a technicality. Naming a lot is exactly what separates a real document from a marketing PDF — and TARE is made by a contract manufacturer and never takes possession of inventory, so nothing connects a published lot to the jar in anyone’s hands. We publish the document, name the lot on it, and do not claim it is yours.

The documents themselves are published. Our manufacturer gave written approval to reproduce them on 17 August 2026, and all three are on Testing & Certification as images — unmarked, uncropped, unredacted. What is still not published is the original PDF: what we hold are the images supplied through the manufacturer’s dashboard.

What a certificate cannot tell you

It cannot tell you whether a formula works. A certificate is a composition and contamination document. It answers what is in the material and how much, and it is silent on every question about outcome. A brand using a certificate as evidence for a benefit is using the wrong document, confidently.

It cannot tell you the material in your jar is the material on the paper. See above. Anyone promising that is promising something a contract-manufactured supply chain does not deliver.

And the test worth running on any brand, including this one: ask for the document, then read the result column. Almost nobody in this category publishes one at all, and nobody publishes one showing they claim less on the label than the assay found. That comparison takes ten seconds and it is the whole of what we are asking anyone to do.

More in this cluster: what a GMP claim actually covers, the format that makes the arithmetic impossible, and the sixty-second label procedure.